fix(bulk_writer): preserve remote upload prefixes on Windows

Signed-off-by: VectorPeak <73048950+VectorPeak@users.noreply.github.com>

Author: VectorPeak

pymilvus/bulk_writer/remote_bulk_writer.py

@@ -294,10 +294,12 @@ def _upload(self, file_list: list):
                 if ext not in [".json", ".jsonl", ".npy", ".parquet", ".csv"]:
                     continue
 
-                relative_file_path = str(file_path).replace(str(super().data_path), "")
-                minio_file_path = str(
-                    Path.joinpath(self._remote_path, relative_file_path.lstrip("/"))
-                ).lstrip("/")
+                relative_file_path = (
+                    Path(file_path).resolve().relative_to(super().data_path).as_posix()
+                )
+                minio_file_path = (
+                    f"{self._remote_path.as_posix().strip('/')}/{relative_file_path}"
+                )
 
                 if self._object_exists(minio_file_path):
                     logger.info(

tests/unit/test_remote_bulk_writer.py

@@ -112,10 +112,13 @@ def test_upload_jsonl_file(self, tmp_path):
             uploaded_files = writer._upload([str(jsonl_file)])
 
         assert len(uploaded_files) == 1
-        assert uploaded_files[0].endswith(".jsonl")
+        assert uploaded_files[0].endswith("/1.jsonl")
         mock_upload_object.assert_called_once()
         _, kwargs = mock_upload_object.call_args
         assert kwargs["file_path"].endswith(".jsonl")
-        assert kwargs["object_name"].endswith(".jsonl")
+        assert kwargs["object_name"] == uploaded_files[0]
+        assert kwargs["object_name"].endswith("/1.jsonl")
+        assert "bulk/test/" in kwargs["object_name"]
+        assert "\\" not in kwargs["object_name"]
         mock_local_rm.assert_called_once_with(str(jsonl_file))
-        assert writer.batch_files[0][0].endswith(".jsonl")
+        assert writer.batch_files[0][0] == uploaded_files[0]